When the attacker has AI too
Defending financial infrastructure in the adversarial AI era
June 10, 2026
Key takeaways
- AI has materially expanded the attack surface for financial services infrastructure, lowering the cost and raising the ceiling for attackers across every stage of the attack lifecycle.
- Sophisticated threat actors now use AI to accelerate phishing, automate vulnerability discovery and exploit generation, orchestrate denial-of-service attacks and probe fraud detection models, increasing the speed and scale of attacks.
- Trusted defensive practices that protected institutions in the past may not scale in the future. Reviewing resiliency plans, patching SLAs and detection capabilities is critical to staying ahead of AI-driven threats.
For the past several years, the dominant conversations around AI in the financial sector have focused on how emerging AI capabilities can be leveraged to improve the efficiency of operations – or sometimes from a risk lens, regarding how AI capabilities may be misconfigured, contain bias or be misused in some way. While these topics remain important, we must also now urgently focus on how sophisticated threat actors, including cybercriminals and nation-states, may leverage AI capabilities to attack our institution.
The position of FIS® Cybersecurity, based on our direct testing and engagement with the broader cybersecurity community, is that the attack surface for financial services infrastructure has expanded materially. The speed and scale of attacks have increased, and the trusted defensive practices that protected our institutions in the past may not scale effectively in the future.
This article lays out what we are seeing and what we recommend our clients do to manage these emerging AI security risks.
24/7
Cyber defense center
Continuous monitoring with agentic AI-driven triage and incident remediation24–48h
Critical CVE patching SLA
FIS target for remediating critical vulnerabilities across all internet-facing systems8
Recommended control areas
Control domains FIS recommends all partner institutions comprehensively reviewHow adversaries are using AI against financial infrastructure
AI has lowered the cost and raised the ceiling for attackers across every stage of the attack lifecycle – from initial attack reconnaissance through data exfiltration and evasion of detection.
AI-driven attack patterns we consider most relevant today:
Threat vector: AI-accelerated phishing and social engineering
Severity: Critical
Impact description: Frontier language models allow attackers to generate highly personalized spear-phishing content – correctly formatted emails, synthetic voice messages and deepfake videos. We have observed campaigns targeted at treasury and payments staff that use publicly available data to identify targets with contextually accurate pretexts.
Threat vector: Automated vulnerability discovery and exploit generation
Severity: Critical
Impact description: Adversarial usage of AI capabilities allows attackers to probe bank financial software platforms to identify weaknesses in codebases, APIs and network configurations. Such capabilities also vastly shrink the window between the discovery of zero-day vulnerabilities and their exploitation against victim institutions
Threat vector: AI-driven transactional fraud
Severity: High
Impact description: Adversarial usage of AI allows an attacker to probe a bank’s fraud detection models – submitting carefully crafted transactions designed to stay below detection thresholds. This is particularly concerning for real-time payments infrastructure, where transaction volumes make manual reviews impractical
Threat vector: Adversarial attacks on AI models leveraged by financial institutions
Severity: High
Impact description: Banks are deploying AI at record pace – including for critical internal and client-facing use cases. These models may themselves become targets, with adversaries attempting to modify their output by poisoning training data, crafting adversarial inputs or launching other AI-centric attacks.
Threat vector: AI-orchestrated denial-of-service attacks
Severity: High
Impact description: AI is being leveraged to orchestrate distributed denial-of-service attacks with greater throughput and adaptiveness – especially when coupled with AI-driven technology enumeration and identification of weaknesses.
While frontier models, including Mythos from Anthropic and comparable limited-access offerings from OpenAI and Google Gemini, have garnered headlines, our research has determined that even the most widely available models today are highly effective at identifying and exploiting security weaknesses. Further, with high confidence we assert that these capabilities will continue to proliferate in the months and years ahead. The following are the AI-driven attack patterns that we consider most relevant to our client institutions:
What your institution should be doing now
Cybersecurity and the protection of our critical financial services platforms is a top priority for FIS. As such, we have spent the past few months iteratively improving our existing cybersecurity and technology management capabilities to stay ahead of increasing threats.
FIS recommends that the following control areas be thoroughly reviewed by our partner institutions:
1. Build and maintain a comprehensive asset register, including AI systems
Defensive posture against adversarial AI starts with knowing your technology footprint. Each system, application, and network that your institution relies on must be inventoried and documented.
- Develop a comprehensive inventory of systems, networks, and applications, including cloud environments, SaaS services, and software provided by third parties. Establish clear ownership for each and assure records are regularly updated.
- Specifically denote systems that are exposed to the internet or to external parties via B2B connections.
- Include specific inventory of AI capabilities, including AI systems developed internally and those sourced from third parties.
- Maintain Software Bill of Material (SBOM) documentation, detailing open and closed-source components composing your firm’s software.
- Review and re-attest to the completeness of this inventory, at least bi-annually; establish processes to automate updates as your technology footprint evolves.
2. Establish an AI-ready patching policy
Traditional approaches to vulnerability management will not be sufficient as AI-driven vulnerability discovery and attacks continue to proliferate. Review patching policies to assure that appropriate mechanisms are in place to evaluate vulnerability exposures, triage risk and remediate with proper urgency. Leverage automation to improve the efficiency of vulnerability discovery and patch deployment.
Optimize practices for managing defects and vulnerabilities within software developed by your organization – including honing capabilities to quickly triage and mitigate exposure to emerging vulnerabilities in open-source components.
Thoroughly review patching SLAs – implement practices now that allow for ‘Critical’ level vulnerabilities to be consistently mitigated or remediated across your externally accessible systems within 24-48 hours.
3. Harden external-facing infrastructure and enforce least-privileged access
Internet-facing systems – applications, APIs and corporate infrastructure – represent the most critical exposure surface and should be thoroughly reviewed and hardened.
- Disabling unnecessary services.
- Assuring that operating systems, software, and other components are up to date.
- Reviewing identity systems to assure that only the least-required privileges are issued for all user accounts, and that privileged and nonhuman identities are sufficiently managed.
- Defense-in-depth controls, including endpoint security platforms, intrusion prevention systems and web application firewalls are properly tuned and configured to automatically update defenses based on the latest threat intelligence.
4. Exhaustively test systems using adversarial testing
Assure that all systems are subject to penetration testing at an appropriate cadence. More exposed systems (i.e. those on the internet) should be tested most frequently, including regular “red team” style testing leveraging the latest attack techniques exhibited by real-world attackers. Leverage AI-driven adversarial testing to expand the depth and breadth of your testing capabilities.
5. Assure detection and monitoring capabilities are AI-ready
Existing security monitoring and response functions should be thoroughly evaluated to assure their capability in maintaining pace with the increased volume and sophistication of cyberattacks. Performance metrics, including handling SLAs, should be closely monitored to assure efficacy in maintaining pace. Monitoring and response workflows also represent great opportunities for implementing agentic AI capabilities to augment human defenders.
Leverage resources from FS-ISAC and other trusted industry sources to assure your firm is ingesting and actioning the latest available threat intelligence and readiness resources – allowing your firms to proactively prepare for increasing AI threats and remain tactically protected day-to-day.
6. Strengthen third-party and supply chain risk management
AI security risks raise the stakes for managing third and fourth-party security risks. Ensure that your firms have effective mechanisms in place to understand supply-chain usage of AI and the capabilities your partners have in place to manage emerging AI-driven security risks.
Add AI-specific security questionnaires into third-party due diligence processes and consider contractual rights for the audit of AI components and cybersecurity capabilities.
7. Collaborate
One of the true beauties of the financial services industry is that no one is left on an island when it comes to managing cyber risks – the collaboration across the sector on both tactical and strategic risks provides each firm with major benefits.
This holds true for managing AI security risks – groups such as FS-ISAC and BPI have bespoke assessment models to help firms understand exactly where they stand on readiness, as well as working groups and other resources to arm firms in best managing these emerging risks.
8. Thoroughly review and test resiliency plans
Review resiliency plans to assure that appropriate procedural and technical controls are in place to achieve designated recovery SLAs. Leverage tabletop testing as a mechanism to evaluate efficacy of response to AI-driven security incidents, including ability to recover critical systems.
How FIS is handling this
At FIS, we recognize the critical role that we play in delivering resilient financial services systems. As such, over time we have worked diligently to assure we have the right level of cyber defenses, as well as leveraged recent risk trends to further increase our cyber defenses.
Our capabilities include:
- 24/7 Cyber Defense Center, including usage of agentic AI capabilities for triage and remediation of incidents.
- Mature cyber hygiene functions, including capabilities to drive rapid assessment of zero-days and other vulnerability exposures and mitigation of the most critical vulnerabilities on externally facing systems within 24-48 hours.
- Augmented existing SDLC and product security practices with scanning of source code and runtime environments leveraging frontier models such as Anthropic’s Mythos.
- Implementation and continuous intelligence enrichment of leading preventative and detective controls within our defense-in-depth stack.
- Enhancement to our technology operations practices, including increasing the cadence and efficiency of security patching.
- Increasing our rigor and diligence in third party oversight, including evaluation of cyber defenses and AI security risks.
- Improving our fraud mitigation capabilities to maintain pace with shifting fraud trends
- Active collaboration with our clients, including via industry groups such as BPI, FS-ISAC and FS-ARC.
Similar Articles
Accelerating digital transformation in capital markets
Drive efficiency and innovation in capital markets with AI, cloud technology and data modernization to meet evolving digital demands.
Cloud computing is a strategy that financial services firms need to get right
Move financial services to the cloud to boost innovation, enhance stability, improve security and streamline operations efficiently.
How AI is transforming the future of banking
AI and data are reshaping banking with innovations in customer experience, fraud prevention, compliance and operational efficiency.